Activist Fiscal Policy

Transcription

1 Journal of Economic Perspectives Volume 24, Number 4 Fall 2010 Pages Activist Fiscal Policy Alan J. Auerbach, William G. Gale, and Benjamin H. Harris During and after the Great Recession that began in December 2007 (according to the Business Cycle Dating Committee at the National Bureau of Economic Research), the U.S. federal government enacted several rounds of activist fiscal policy. These began early in the recession with temporary tax cuts enacted in February 2008, followed by a tax credit for first-time homebuyers enacted in July They reached a crescendo in February 2009 with the American Recovery and Reinvestment Tax Act (ARRA): a combination of tax cuts, transfers to individuals and states, and government purchases estimated to increase budget deficits by a cumulative amount equal to 5.5 percent of one year s GDP. The fiscal stimulus continued thereafter with more targeted measures, notably the temporary cash for clunkers program in summer 2009 aimed at stimulating the replacement of old cars with new ones, and an extension and expansion of the First-Time Home- buyer Credit in November 2009 and July Accompanying these fiscal efforts were the Troubled Asset Relief Program, enacted in fall 2008 to address the finan- cial crisis, and a continuing array of interventions by the Federal Reserve Board that aimed to stabilize credit markets and stimulate the economy. Around the world, other countries caught in the grip of recession also pursued a variety of active fiscal strategies, ranging from temporary consumption tax rebates (for example, in the United Kingdom) to large public works projects (notably in Alan J. Auerbach is the Robert D. Burch Professor of Economics and Law, University of California, Berkeley, California. William G. Gale is the Arjay and Frances Miller Chair in Federal Economic Policy, The Brookings Institution, Washington, D.C. Benjamin H. Harris is Senior Research Associate, Economic Studies Program, The Brookings Institution, Washington, D.C. Their addresses are auerbach@econ.berkeley.edu, wgale@brookings.edu, and bharris@brookings.edu. doi= /jep

2 142 Journal of Economic Perspectives China). The prevalence of fiscal policy interventions in this period reflects both the severity of the recession and a revealed optimism with regard to the potential effectiveness of activist fiscal policy. Yet the variety of policies adopted also suggests uncertainty about which approaches might have been most effective. In this paper, we review the recent evolution of thinking and evidence regarding the effectiveness of activist fiscal policy. Although fiscal interventions aimed at stimulating and stabilizing the economy have returned to common use, their efficacy remains controversial. We review the debate about the traditional types of fiscal policy interventions, such as broad-based tax cuts and spending increases, as well as more targeted policies. We conclude that while there have certainly been some improvements in estimates of the effects of broad-based policies, much of what has been learned recently concerns how such multipliers might vary with respect to economic conditions, such as the credit market disruptions and very low interest rates that were central features of the Great Recession. The eclectic and innovative interventions by the Federal Reserve and other central banks during this period highlight the imprecise divisions between monetary and fiscal policy and the many channels through which fiscal policies can be implemented. The Fall and Rise of Activist Fiscal Policy Until very recently, a typical student of macroeconomics would likely be introduced to discretionary fiscal policy through a cautionary tale of the hubris of attempts at fine tuning in earlier decades. The student would start with a group of well-rehearsed arguments, beginning with the lags in the making of economic policy and further lags in the implementation and effects after the policy is enacted, which make it difficult for policymakers to time fiscal policy actions to stabilize the economy. Indeed, a recession could end even before the need for action was recog- nized, with government officials still focused, as they were in 1975, on the need to Whip Inflation Now. The student would also learn that uncertainty about policy multipliers made weaker intervention desirable (Brainard, 1967). The student would learn the Lucas (1976) critique, which implies that a policy s stabilizing effects can be undercut by the expectations and actions of rational agents who observe the government s policy process. For example, one reason that investment might drop during a recession is the anticipation that a countercyclical investment incentive will be enacted in the near future. Consumption might not respond much to a counter- cyclical reduction in income taxes, because the wealth effects of such tax reductions are small when the reductions are seen as temporary. The intriguing notion of Ricardian equivalence (Barro, 1974) would promote further skepticism about the effectiveness of fiscal policy. Finally, the student would be reminded of the alternative tools of stabilization policy, notably the interest-rate interventions of independent central banks and the automatic stabilizers already built into the government s tax and transfer systems. Indeed, prior to 2008, the student would probably learn that,

3 Alan J. Auerbach, William G. Gale, and Benjamin H. Harris 143 through such alternative interventions, a Great Moderation in postwar economic performance had been achieved. 1 This array of arguments against activist fiscal policy clearly met its match during the Great Recession, when those policymakers not already imbued with the Keynesian doctrine rediscovered the old-time religion in their foxholes. But it is not accurate to say that activist fiscal policy was totally discredited or unpracticed in the period just before. In the United States, a resurgence in fiscal policy intervention is clearly detectable in the last decade. As shown in Auerbach and Gale (2009a), simple policy reaction functions, measuring the legislated responses of federal taxes and spending to the state of the economy and the budget, show evidence of much stronger responses to both factors, particularly to the economy, in the period from the start of the George W. Bush administration through the 2007 turning point relative to the three previous presidential administrations. This recent increased countercyclical policy activism is nicely highlighted by the very different policy responses during the two recessions before the most recent two. In August 1982, after a year of deep recession that still had several months left to run, Congress passed the Tax Equity and Fiscal Responsibility Act (TEFRA), scaling back the large Reagan tax cuts that had been enacted just over one year earlier. Legislation over the same period cut near-term federal spending. During the next U.S. recession, in October 1990, a budget summit meeting of President Bush and Congressional leaders produced legislation aimed at reducing the deficit. Thus, in 1982 and 1990, policymakers chose to impose fiscal discipline during a recession. This pattern changed in the 2000s. In 2001, as concerns about a recession developed, Congress added a set of cash rebates to the original set of proposed Bush tax cuts in order to help stimulate the economy in the short run. In early 2002, in response to the 2001 recession that was not then known to have ended, Congress introduced bonus depreciation, the first use of countercyclical investment incen- tives since the 1970s. In 2003, further individual tax rebates were enacted, as part of a package that focused mainly on other changes. Early 2008 saw the first round of fiscal stimulus during the Great Recession, adopted just two months after the expan- sion was later determined to have ended and at a time when few economic forecasts predicted a deep recession. For example, the Congressional Budget Office (2008b) economic outlook for 2008 and 2009 released in March 2008 forecasted real GDP growth rates of 1.9 percent and 2.3 percent and unemployment rates of 5.2 and 5.5 percent, respectively. While some of the explanation for this quicker and more sustained resort to fiscal policies may lie in the relaxation of budget rules, which made countercyclical fiscal interventions easier (Auerbach, 2008) and some 1 Stock and Watson (2002) argue that the decline in economic volatility can be attributed to a mix of a more aggressive Federal Reserve policy towards inflation, less volatile productivity and commodity price shocks, and certain unknown factors. Kahn, McConnell, and Perez-Quiros (2002) and Davis and Kahn (2008) attribute decreased volatility to improved inventory management, especially in the durable goods sector; Davis and Kahn (2008) find no corresponding decline in wage, income, and household consumption volatility.

4 144 Journal of Economic Perspectives may lie in the politics of tax cuts and their support by the Bush administration, developments in economic theory and evidence had also provided a stronger foun- dation for at least some discretionary policy interventions. Fiscal Models and Fiscal Multipliers Besides the timing of fiscal changes, discussed above, the strength of activist fiscal policy is a central issue regarding such interventions. The effect of policy is typically measured via a multiplier. The multiplier is the ratio of the rise in GDP rela- tive to the size of the policy intervention (the reduction in taxes and/or increase in government purchases), with both terms defined more carefully below. A multiplier of 1 means that GDP rises by the size of the fiscal intervention. A multiplier greater than 1 means the economy grows by more than the stimulus. A multiplier between 0 and 1 indicates that the economy grows, but by less than the actual stimulus. While a larger multiplier is, of course, a better outcome when a policy is aimed at increasing economic activity, a positive multiplier of any size indicates that the policy raises GDP. For a tax cut or an increase in transfer payments (which do not alter GDP directly), the multiplier represents the increase in both the aggregate economy and private sector activity. For government purchases, the increase in private sector activity per dollar of government purchases equals the multiplier minus one. Thus, a multiplier of less than 1 for an increase in purchases would indicate that some private sector activity is being crowded out. 2 In any analysis, it is important to clarify the definition of the multiplier employed, since both the size of the policy intervention and the effect on GDP vary over time for most policies. Some studies relate the cumulative change in GDP to the cumulative change in taxes or spending over some relevant term, typically five years or less, while others relate the peak change in GDP to the peak change in the policy variable, with the most natural definition somewhat dependent on the timing and duration of the policy intervention. There is no single right way to perform the calculation, and qualitative comparisons across policies and studies sometimes fail to specify the exact multiplier concept used. The effects of fiscal policy can usefully be divided into direct effects and economywide effects. For some policies, such as the rebates introduced earlier in the decade, data at the individual level can be used to estimate responses. Similar approaches can be used to estimate the effect of tax incentives on investment, although this line of research has proved challenging for several reasons. We review some estimates from both of these literatures in some detail below. These approaches, however, only estimate the direct responses to tax changes and not the 2 The discussion above refers to tax cuts or spending increases, in which case a positive multiplier indicates a positive effect on GDP. As discussed later in the paper, there is a possibility that fiscal consolidation that is, a cut in government purchases or an increase in taxes could boost GDP, in which case the resulting multiplier would be negative.

5 Activist Fiscal Policy 145 effects on economywide activity, which could be smaller or larger than the direct effects. As a result, we review a variety of models that take account of the various additional channels through which tax cuts, transfers to individuals and states, and increases in government purchases affect GDP and its components. Direct Effects Tax cuts to stimulate consumption have a long history. These policy efforts have generated a substantial literature, reviewed in greater detail in Auerbach and Gale (2009a), that offers several fairly robust results about the marginal propensity to consume (MPC) out of tax cuts. First, consistent with standard life-cycle and permanent-income models, most of the evidence suggests that household consumption responds more vigorously to tax changes that are plausibly expected to be longer-lasting than to changes that are expected to be shorter-lasting, with estimates of a MPC of 0.9 for long-lived poli- cies. Second, household responses to a given tax cut are heterogeneous. As theory predicts, borrowing-constrained households tend to have a larger MPC out of tax cuts than do other households, and low- and middle-income households are more likely to be borrowing-constrained than upper-income households. Third, the effect of tax changes on consumer spending tends to occur when the policy change is implemented, not when it is enacted or credibly announced. 3 While these three findings are generally consistent with standard optimizing behavior in a setting where some households face borrowing constraints, other results suggest the importance of an additional set of factors namely, the way tax cuts are described and delivered. These results are consistent with a growing literature indicating that framing, presentation, and other factors, such as default specifications, have a significant influence on saving behavior, and therefore are relevant because saving and consumption choices are closely linked. For example, some evidence from survey data suggests that adjustments to tax withholding that do not represent tax cuts can nevertheless affect consumption (Shapiro and Slemrod, 1995). Households appear to adhere more closely to standard model predictions when the policy-induced changes in income are large (Hsieh, 2003). Comparing estimated marginal propensities to consume for the 2001 and 2008 tax cuts provides interesting perspectives on two issues noted above the role of tax cut permanence and of heterogeneous responses. The 2001 rebate was clearly even at the time of enactment part of a longer-lasting tax cut, whereas the 2008 rebate was explicitly a one-time event. On the other hand, the 2001 rebate 3 For evidence on the marginal propensity to consume from short-lived policies, see for example Blinder (1981), Blinder and Deaton (1985), and Poterba (1988); for corresponding evidence on the marginal propensity to consume from longer-lived policies, see Souleles (2002) and Johnson, Parker, and Souleles (2006). For evidence on the links between borrowing constraints and a higher marginal propensity to consume, see for example Johnson, Parker, and Souleles (2006), Broda and Parker (2008), Agarwal, Liu, and Souleles (2007), and Bertrand and Morse (2009). For evidence that the policy effects occur after implementation, see for example Campbell and Mankiw (1989), Wilcox (1989), Parker (1999), Souleles (1999, 2002), Poterba (1988), Blinder and Deaton (1985), Johnson, Parker, and Souleles (2006), Broda and Parker (2008), and Watanabe, Watanabe, and Watanabe (2001).

6 146 Journal of Economic Perspectives went to all income groups and was not refundable, whereas the 2008 rebate was limited to low- and middle-income households and was refundable. (A refundable tax rebate is available in full even to individuals who owe no taxes or whose liability is smaller than their refund amount.) The first difference should raise the MPC out of the 2001 rebate relative to 2008; the second difference should reduce it. In fact, estimated MPCs are not significantly different for the two tax cuts. For example, Broda and Parker (2008) examine micro data on household purchases and find that households consumed about 20 percent of the rebate in the first month after receiving it, a rate of consumption that is consistent with the MPC out of the 2001 tax cuts reported in Johnson, Parker, and Souleles (2006). Shapiro and Slemrod (2003, 2009) report the results of asking respondents in phone surveys how they intended to use the 2001 and 2008 tax cuts, respectively, and report a remarkable similarity in overall responses for the two tax cuts. For example, 21.8 percent of households said they would mostly spend the 2001 tax cut, compared to 19.9 percent for the 2008 tax cut. The literature on the effect of federal transfers on consumption is not as extensive as analysis of tax cuts, but it shows clearly that transfer payments do affect household consumption. Gruber (1996, 1997) demonstrates strong effects on contemporaneous consumption from increases in welfare payments and unemployment insurance bene- fits, respectively. Edwards (2004) estimates a marginal propensity to consume out of Earned Income Tax Credit payments of approximately 0.7. Barrow and McGranahan (2000) also find strong effects of EITC receipts on spending. Several studies examine the responsiveness of business fixed investment to changed investment incentives. But estimating investment responses is a consider- ably more challenging exercise, for at least two reasons. First, there are relatively few natural experiments providing changes in investment incentives; there were essentially no changes in the tax treatment of investment between 1986 and Second, investment decisions are more difficult to model, in part because of the interaction of different tax provisions (notably those that affect a firm s financial policy and that limit the ability of firms to utilize tax deductions). A series of studies has focused on the effects of tax changes on the composition of business fixed investment, primarily using panel data on firms, industries, or asset categories (for example, Auerbach and Hassett, 1991; Cummins, Hassett, and Hubbard, 1994; Hassett and Hubbard, 2002). These studies provide ample evidence that changes in the user cost of capital as first defined by Dale Jorgenson as the implicit rental cost of a capital investment that establishes its break-even marginal product do influence the mix of investment, with the elasticity of equipment investment with respect to the user cost of capital falling in a range between 0.5 and 1.0. Using a related methodology, House and Shapiro (2008) estimate invest- ment responses to the bonus depreciation incentives of , finding that the composition of investment did shift from nonqualifying investment to qualifying investment. One interesting result in the House Shapiro analysis is that investment responses to the 2002 introduction of bonus depreciation appeared to begin during the last quarter of 2001 and the first quarter of 2002, a period ultimately covered

7 Alan J. Auerbach, William G. Gale, and Benjamin H. Harris 147 retroactively by the 2002 legislation. Thus, firms may have expected that investment incentives would be enacted and that investment undertaken during this interval would be covered. This predictability of investment incentives should not be particu- larly surprising, given how well one can predict their timing using a relatively simple model (Auerbach and Gale, 2009a), but it can be a cause for concern, given that the effect of announcing a new future investment incentive will tend to reduce current investment, at least if retroactive application of the incentive is not also anticipated. In summary, tax incentives affect investment, with the compositional effects more easily identified than the aggregate effects. But relatively little attention has been given to the announcement effects of policy. Also, it is worth keeping in mind that the conditions governing investment in a recession, such as cash-flow constraints and business losses for tax purposes, may produce quite different invest- ment responses to temporary tax cuts than would be predicted using models based on responses to long-term tax reforms adopted under more normal circumstances. Besides cutting taxes or transferring funds to households and businesses, federal policy can also influence aggregate activity by altering state and local spending and tax policy. This is, in principle at least, a potentially powerful avenue for stimulus, given the magnitude of state and local spending and taxes (more than 12 percent of GDP in 2009) and the fact that almost all states have balanced budget rules. When revenues fall during a recession, states can either draw down their rainy day funds, raise taxes, or cut spending and the latter two options are likely to act as procy- clical policies that could exacerbate the downturn. Poterba (1994), for example, finds strong evidence that states contract spending and raise taxes when faced with a negative fiscal shock. In such cases, federal transfers could ease the constraint and reduce the need for contractionary state responses. While the argument for transfers to states being stimulative is plausible, there is surprisingly little evidence on the countercyclical effects of federal transfers to states. Gramlich (1978, 1979) and Reischauer (1978) evaluate the effects of three federal grant programs undertaken in response to the recession. One program offered countercyclical revenues to the states in the form of block grants, another paid the salaries of state and local government workers, and a third contrib- uted funding for capital improvements. The general finding was that the short-run response by states to federal aid was primarily to bolster state rainy-day funds, with only modest increases in outlays and reductions in taxes in the short run. The long-run response particularly in the form of decreased income tax revenue was substantial, but materialized after the recession had ended. It is unclear how relevant these findings are to the current economic downturn, however, given the dated nature of the evidence, the differences in the states economic situations now (when they have been hurt by both the recession and the housing crisis, which heightened the need for state transfers to local governments due to reduced municipal property tax revenues) and differences between the 1975 economy and the current one. Although the effects of fiscal policy on individual components of output are of interest, and show the responsiveness of particular sectors to fiscal interventions,

8 148 Journal of Economic Perspectives they do not capture the effects on overall output, since they omit the indirect, economywide responses. Economywide Estimates Generally, three types of models have been used to examine the overall economic effects, with differing strengths and weaknesses: large-scale macroeco- nomic models, structural vector autoregressions, and dynamic stochastic general equilibrium models. Large-scale macroeconomic models account for relevant prices and quantities in different sectors of the economy, and relate these prices and quantities to each other and to government policy variables. While large-scale models provide consid- erable detail regarding the channels through which policy can operate, and are commonly used by government forecasters, their theoretical grounding has been challenged based on the argument that the structural equations describing the behavior of households and firms lack adequate microfoundations (Lucas, 1976). Of the three types of models, large-scale macro models often produce the largest multipliers. We discuss results from several large-scale models in subsequent sections when we address the effects of the American Recovery and Restoration Act of The two remaining types of models, which we now consider in turn, have been the mainstays of the recent academic literature. They represent alternative responses to the criticisms of large-scale models. One approach dynamic stochastic general equilibrium (DSGE) models hews more closely to micro-foundations; the other structural vector autogression (SVAR) models moves away from attempts to establish strong structural restrictions and relies to a greater extent on time series methods. In a standard vector autoregression, a vector of variables say, output, taxes, and government purchases is regressed on lagged values of the same variables. Because there is no specification of the channels through which policies affect output, it is not possible to separate the response of output to policy from the response of policy to output. In a structural vector autoregression, a limited structure is provided in the form of assumptions about the order in which shocks to policies and output occur (in more formal terms, assumptions about the recursive structure of the error matrix). These assumptions make it possible to identify the changes in current policy variables that are attributable to actual changes in policy rather than to endogenous responses to economic conditions. The key issue in this literature is the method used to identify true policy changes in attempting to obtain persua- sive multiplier estimates. An important early contribution in the structural vector autoregression litera- ture, by Blanchard and Perotti (2002), provides estimates of multipliers for both government purchases and taxes using the identifying assumption that these vari- ables could respond to output within a quarter (the period of observation) only through automatic provisions, not discretionary policy. Thus, controlling for such automatic response, which could be estimated directly, the fiscal shocks within a period could be treated as exogenous. Based on such a methodology, Blanchard and Perotti estimate a GDP multiplier for government purchases of about 0.5 after

9 Activist Fiscal Policy 149 one year, with longer-term multipliers depending on model specification due to differences in the estimated permanence of policies. That is, the short-term multi- pliers imply a net crowding-out of components of GDP other than the government purchases themselves. Estimates of tax cut multipliers are slightly larger, closer to 1.0 after one year. As noted, a central concern with the structural vector autoregression approach is the identification of policy shocks. A change in taxes or spending identified by the Blanchard and Perotti (2002) methodology as a policy shock might have been anticipated by individuals (even if not by the econometric model) or it might not have been a policy change at all (for example, it might be due to other factors such as a change in the income distribution). Thus, one line of research extending this approach has been to identify policy changes through a narrative approach, applying additional information on policy decisions to help identify exogenous policy changes, rather than treating as exogenous surprises those changes not predicted by the structural vector autoregression itself. Using military spending build-ups as an important source of variation in government purchases that is exogenous with respect to economic activity, Ramey and Shapiro (1997) estimate the effect of these build-ups on GDP and its other components. More recently, Ramey (2009) provides a more complete set of data on such shocks and emphasizes the importance of distinguishing the announcement dates of policy changes from their dates of implementation. Using such a series based on actual policy announcements, she estimates an output multiplier after four quarters of about 0.7. As noted above, one implication of a multiplier below 1.0 for government purchases is that other components of GDP fall in response to the increase in government purchases. On the tax side, the narrative approach to identifying policy shocks has been introduced by Romer and Romer (2007), who used the same approach in earlier analysis identifying monetary policy shocks. They argue that the multipliers of tax changes estimated using other approaches are likely to underestimate tax policy multipliers by treating as exogenous many policy changes that were actually responding to economic conditions or government purchases. Using their narra- tive approach to identify policy changes that were arguably independent of such other factors, they find a GDP tax-cut multiplier of about 1.0 after four quarters, rising to 3.0 after 10 quarters. This very large multiplier is associated with an enor- mous impact on investment. The result is striking: indeed, so striking that it merits further investigation. 4 Although the narrative approach may yield better estimates of true policy surprises than the standard structural vector autoregression approach, both approaches are limited in certain critical respects stemming from the reduced-form 4 For example, Favero and Giavazzi (2009) suggest that the multipliers for the tax shocks identified by Romer and Romer are considerably smaller if one models the shocks as explanatory variables in a multivariate model rather than simply regressing output on the tax shocks. The source of this difference is not clear, although the authors suggest that their results reject the assumptions by Romer and Romer that such shocks are independent of other explanatory variables.

10 150 Journal of Economic Perspectives nature of these models. First, the models cannot be used to examine the economy s responses to automatic stabilizers or to any already-operating rules that relate activist fiscal policy to economic conditions, because effects of both types are already incorporated in the model s estimated impulse responses. Second, these models can measure only the multipliers of policies that deviated from standard policy responses to economic conditions within the sample period and can only estimate the effects of those policies as they were actually adopted. For example, if shocks to government purchases or taxes tended to be short-lived, then we cannot draw direct inferences about the effects of more permanent shocks. New tax changes differing in composition from those examined in-sample could well have different multipliers than those estimated. This concern is especially important under the narrative approach, in the light of the fact that most of the estimates of the effects of government purchases actually relate to defense spending and are based heavily almost exclusively on the experience during World War II or the Korean War (Hall, 2009). Third, these models can only estimate the effects of policy interventions under the economic conditions prevailing within the sample, and the multiplier effects of different policies could vary substantially with economic condi- tions. Investment incentives that might be strong in a boom might be ineffectual in a period of tight credit and net operating losses. Tax cuts for households might have a larger effect during periods in which liquidity constraints bind more tightly. Government spending might have larger multipliers during periods, like recent times, when the zero-interest-rate bound is binding. As a consequence, much of the recent discussion and debate surrounding the potential effects of policy intervention have been based on the analysis of the third approach alluded to above: dynamic stochastic general equilibrium models. These models typically feature a relatively small number of equations based tightly on microeconomic theory, with some parameters derived from empirical estimates and others calibrated to make the model consistent with observed macroeconomic rela- tionships. Because these models specify a full economic structure, they can be used to analyze policies and policy environments in a way that is not limited by historical experience. For example, they can explore interactions between monetary and fiscal policy, the role of long-term fiscal shortfalls on the effect of current stimulus packages, the role of different degrees of openness in the economy, the role of anticipations of fiscal policy actions, and so on. But to do these things, the dynamic stochastic general equilibrium approach leans heavily on modeling assumptions that may or may not be valid: for example, assumptions regarding the stickiness of wages and prices, the prevalence of liquidity constraints, the rationality of agents, the structure of markets, and so forth. Indeed, some of the recent disputes regarding the potential effects of fiscal policies can be traced to differences in the assumptions in dynamic stochastic general equilibrium models as well as to assumptions about the nature and timing of the policies themselves. In a recent review of the dynamic stochastic general equilibrium literature and using his own model of this type, Hall (2009) concludes that plausible dynamic

11 Alan J. Auerbach, William G. Gale, and Benjamin H. Harris 151 stochastic general equilibrium models of the new Keynesian variety (that is, incorporating certain nominal rigidities in wages and prices) generate govern- ment spending multipliers that are consistent with those found using time series methods well above zero, but below 1.0. However, as Hall notes, it appears that in the dynamic stochastic general equilibrium approach, relatively small changes in parameter specification within empirically plausible ranges are capable of producing substantial shifts in estimated multipliers. For example, several recent analyses using dynamic stochastic general equilibrium models, notably papers by Eggertsson (2008) and Christiano, Eichenbaum, and Rebelo (2009), have argued that when nominal interest rates are close to zero, the government spending multi- plier can be substantially larger, with estimates in the range of 3 to 4. 5 One apparent explanation for the larger multiplier under the zero bound is that monetary policy responses are no longer active. The typical dynamic stochastic general equilibrium model includes a Taylor (1993) rule for monetary policy: that is, a rule in which interest rates respond to the output gap and the inflation rate. In normal circumstances, a government spending increase would stimulate output and inflation, which in turn would lead to an increase in interest rates, which would reduce current consumption and investment demand. However, when nominal interest rates fall to the zero bound, this response would be absent, and the output response therefore would be larger, because the monetary authority would still wish for the nominal interest rate to be even lower. This intuition is apparently too simple, though, because some other dynamic stochastic general equilibrium analyses assuming constant interest rates deliver much smaller government spending multipliers. In particular, Cogan, Cwik, Taylor, and Wieland (2009) estimate the response to a permanent increase in government spending, assuming that interest rates stay equal to zero for the first two years of the experiment and follow a Taylor rule of reacting to unemployment and infla- tion thereafter. They find an original multiplier around 1, but that by the end of the two-year period, the effect on output is only 0.4. They attribute this difference from papers finding larger multipliers to a shorter zero-bound period. This finding is consistent with the analysis presented by Woodford (2010) that multipliers are reduced to the extent that the increase in government spending extends beyond the end of the zero-bound period. Thus, the multiplier for government purchases would be largest for a temporary spending increase that extended only for the period in which the interest rate was near the zero lower bound. Another factor that might influence fiscal multipliers is the government s long-term fiscal position. There are many reasons to think fiscal policies would have different effects if they are adopted during a period of fiscal stress than they would otherwise. An extensive theoretical and empirical literature argues that 5 Although these models are more sophisticated, they echo the logic of simpler Keynesian models regarding the effectiveness of expansionary fiscal policy in a liquidity trap. Eggertsson (2008) also argues that a tax cut would be less expansionary in the zero-bound case, in fact having a negative effect on output, because its positive supply-side effects could have deflationary consequences. But this conclusion would only apply to tax cuts that affected marginal tax rates.

12 152 Journal of Economic Perspectives contractionary fiscal policy adopted during periods of budget stress can even have an expansionary effect on output, essentially by shifting the economy s trajectory away from one that could be very constraining for productive activity because of high marginal tax rates or economic disruptions (Giavazzi and Pagano, 1990; Alesina and Ardagna, 1998; Alesina, Perotti, and Tavares, 1998). The empirical evidence, based on panel data for OECD countries, does suggest that fiscal consolidation has a less- contractionary effect when adopted under fiscal stress, as measured by high debt and projected government spending relative to GDP (Perotti, 1999). Analysis based on OECD data also indicates that fiscal contractions are more expansionary when implemented through cuts in government spending, as one might expect given the potential damage from reliance on higher marginal tax rates (Ardagna, 2004). One channel through which the differing effects of fiscal policy under different initial conditions may occur is through expectations of how the deficit resulting from a stimulus will be closed in the future. Several recent papers utilizing the dynamic stochastic general equilibrium modeling approach address this issue with mixed results (Corsetti, Meier, and Muller, 2009; Davig and Leeper, 2009; Leeper, Walker, and Yang, 2009). In summary, while the different approaches used to model and analyze the direct and indirect effects of economic stimulus options have improved significantly in recent years, the literature nevertheless shows a substantial amount of variation in key results. Coenen et al. (2010) represents a noteworthy effort to systematize and under- stand these quantitative differences, using dynamic stochastic general equilibrium models that are employed at the Federal Reserve Board, the European Commission, the International Monetary Fund, the Bank of Canada, the European Central Bank, and the OECD. The American Recovery and Restoration Act of 2009 The American Recovery and Restoration Act of 2009 (ARRA, Public Law 111-5) can be viewed as the continuation of a series of activist fiscal policy interventions dating back to But ARRA was of a different scale than previous efforts. The direct cost of the bill (excluding interest payments on accumulated debt) was originally estimated to be $787 billion over 10 years (Joint Committee on Taxation, 2009) and later revised to $862 billion (CBO, 2010). 6 The policies were to be phased in over time, with $200 billion occurring in fiscal year 2009, $404 billion occurring in fiscal year 2010, and the remainder occurring in fiscal year 2011 or afterwards. Table 1 summarizes the major provisions of the bill and CBO s (2009a) range of estimates of the multipliers associated with each item. In broad terms, 6 Most of the $75 billion increase in the estimated cost of the bill in CBO (2009a) was attributed to higher projected outlays, including an additional $21 billion for unemployment insurance, $34 billion more for the Supplemental Nutrition Assistance Program, and an extra $26 billion for the Build America Bond program; relatively small changes in the projected cost of other initiatives account for the remainder of the difference (CBO, 2010).

13 Activist Fiscal Policy 153 Table 1 Estimated Impact of the American Recovery and Reinvestment Act of 2009 on Output and the Budget, Estimated policy multiplier Category High Low 11-year budgetary cost of provisions (billions of dollars) Federal government purchases of goods and services Transfers to state and local governments for infrastructure Transfers to state and local governments not for infrastructure Transfers to individuals One-time payments to retirees Two-year tax cuts for lower- and middleincome individuals One-year tax cuts for higher-income individuals Extension of first-time homebuyer credit Business tax provisions Source: Congressional Budget Office (2009a). Notes: As reported by the CBO, the policy multiplier is the cumulative impact on GDP over several quarters of various policy options. This table includes provisions scored by the CBO or the Joint Committee on Taxation as totaling $5 billion or more in budgetary costs over the period. Selected provisions with lower total budgetary costs were included if the cost in the period was large. Costs do not add up to the total budgetary cost of $787 billion presented in CBO s cost estimate because several provisions are excluded (because CBO s analysis of those provisions cannot easily be summarized by a single multiplier) and because the costs listed are translations of the budgetary costs to categories of the national income and product accounts. the provisions can be divided into tax cuts, assistance to states and individuals, and investments. The two largest tax cuts were the Making Work Pay Credit and the one-year extension of the higher Alternative Minimum Tax deduction. The Making Work Pay Credit is a refundable tax credit of up to $400 per taxpayer ($800 for couples), equal to 6.2 percent of earned income for 2009 and 2010, with the value of the credit phasing-out for individuals with higher incomes. The stimulus package also expanded the eligibility criteria and raised the maximum value of the Earned Income Tax Credit, expanded the refundability of the Child Tax Credit, and created the American Opportunity Tax Credit which replaced the Hope Credit and expanded tax incentives for higher education. Smaller provisions in the stimulus package included a revised tax credit for the purchase of a new home, suspension of the taxation of unemployment benefits, and a deduction for sales tax paid on the purchase of a new car. Tax cuts for businesses were small relative to tax cuts for individuals, but include an extension from two years to five years in the amount of time that small businesses

14 154 Journal of Economic Perspectives could carry back net operating losses to offset taxable income. The American Recovery and Restoration Act also increased the amount of subsidized bonds that local governments can issue for private activity in economically-distressed areas. Altshuler et al. (2009) describe and evaluate the tax provisions contained in the stimulus package. A substantial portion of the American Recovery and Restoration Act provided aid to individuals and transfers to states, mainly through Medicaid and other programs administered by the Department of Health and Human Services, unem- ployment compensation, and food stamps. Transfers to the State Fiscal Stabilization Fund, a mechanism for providing education funding to states, were also significant, as were one-time economic recovery payments to Social Security beneficiaries, veterans, and individuals receiving Supplemental Security Income. A primary objective of the stimulus package was to increase funding for public infrastructure programs. The major investments revolved around renewable energy, health care research, health information technology, subsidized infrastructure financing, and education programs such as Pell Grants. Amounting to 5.5 percent of current-year GDP, albeit spread over several years, the American Recovery and Restoration Act was the largest stimulus package in modern U.S. economic history. Romer (2009) notes that the largest stimulus provi- sion during the Great Depression amounted to 1.5 percent of GDP and was followed one year later by deficit reduction policies. By way of comparison, almost all OECD countries have introduced stimulus measures, with the packages averaging 2.5 percent of GDP. Automatic stabilizers, however, are substantially smaller in the United States than in most other OECD countries. As a result, while the United States had the largest discretionary stimulus package, the combined effects of its automatic and discretionary policies on the government s budget for were the sixth largest as a share of GDP in the OECD (OECD, 2009). Although the 2009 stimulus package was adopted during a period of very weak economic performance, it encountered criticism on several fronts. The criticisms can largely be summarized by asking whether the package was in a phrase used by Lawrence Summers (2007) sufficiently timely, targeted, and temporary. First, there was concern that the policies, although signed into law in February 2009, would be implemented only gradually, with much of the effect coming after the recession was over and the recovery underway. Of course, this concern about policy lags is one of the standard criticisms of countercyclical fiscal policy. However, it seems somewhat less relevant in the present context, if projections of a long and slow recovery are to be believed. Figure 1 shows the path for GDP relative to potential as projected in March 2009 by the Congressional Budget Office, for the baseline without the February 2009 stimulus package and for two scenarios with the fiscal package, corresponding to CBO s perceived range of multiplier estimates for the package s different components. (These multipliers are detailed in Table 1.) Under these projections, the economy would not reach its potential GDP until 2014, and the stimulus package with three-quarters of its effects taking place in the

15 Alan J. Auerbach, William G. Gale, and Benjamin H. Harris 155 Figure 1 Estimated Impact of 2009 Fiscal Stimulus 1 0 Percent of potential GDP Low effect High effect Baseline Fourth quarter of calendar year Source: Congressional Budget Office (2009a). Note: Figure 1 shows the path for GDP relative to potential as projected in March 2009 by the Congressional Budget Office for the baseline without the February 2009 stimulus package and for two scenarios with the fiscal package, corresponding to CBO s perceived range of multiplier estimates for the package s different components. first 18 months would speed the rate of approach. Thus, while more rapid imple- mentation might have been preferred, the biggest avoidable delay was probably at the enactment stage that is, the time period late in 2008 in which a lame-duck President Bush and the outgoing Congress deferred actions for months even after the likelihood of intervention became high. The risk of destabilizing the economy by injecting fiscal stimulus into an overheating economy seems to be less of an issue. The desire to keep the package temporary is motivated by concerns about the long-term budget outlook. However, the stimulus package contributed less to the current-year deficit than did the recession itself, through automatic stabilizers working primarily on the tax side. As we have discussed elsewhere, the contribu- tion of the stimulus package to the long-term U.S. fiscal problem is minimal, if one assumes that the provisions of the stimulus are temporary as enacted (Auerbach and Gale, 2009b). The inclusion in the stimulus package of a number of provisions designed orig- inally without the recession in mind including some of the investments described above highlights the third set of concerns: that the package may not have been well-targeted to provide the strongest fiscal stimulus per dollar of revenue loss or spending increase. Some critics focused on the composition of the package, ques- tioning whether projects that were shovel-ready were likely to be of high value to society and whether the particular tax cuts adopted were the right ones from

16 156 Journal of Economic Perspectives a longer-term perspective. While the stimulus package was certainly not as well- targeted as it could have been, there was some logic to its structure. As noted, the package contained substantial tax cuts, aid to states and individuals, and government investments. The tax cuts should stimulate aggregate demand, but could have been designed more effectively. The aid to individuals was based on humanitarian needs. The aid to states was based on the notion, noted above, that because essentially all states adhere to some form of balanced-budget rule, economic declines that reduce state revenues force cuts in state spending. From the perspective of macroeconomic stabilization, reducing public spending during a sharp downturn is counterproduc- tive. The aid provided should offset some of the state and local spending cuts that would otherwise have occurred. The fact that state and local government spending and employment rose in the second quarter of 2009 is consistent with the view that the transfers supported and stabilized state budgets. In addition, because much of the aid to states was based on criteria such as Medicaid eligibility which is a means-tested program and state unemployment rates, the transfers to states were somewhat targeted to regions most in need of stimulus. Government investments were part of a longer-term Obama administration agenda and are probably not best evaluated as stimulus measures. How well-targeted the package was and the size of the resulting policy multi- pliers remain an area of controversy. 7 Even before the stimulus package was adopted in February 2009, the Obama administration released a document written by Bernstein and Romer (2009) estimating the effect of a potential stimulus plan on employment. These projections were based on estimates of multipliers for govern- ment purchases and tax cuts averaged over those from the Federal Reserve s FRB/ US model and a private forecasting model. The resulting multiplier for a permanent change in government purchases was about 1.5, reached after about one year; the corresponding multiplier for tax cuts (other than investment incentives) was about 1.0, with about three-fourths of the impact reached after one year and the full effect reached after two years. These multipliers are consistent with those assumed by the Congressional Budget Office (2009a, Table 1) in making its projections, in that both the government-spending multiplier and the tax-cut multiplier fall roughly midway between the upper and lower bounds CBO lists for its high-multiplier and low-multiplier scenarios. The similarity in multiplier assumptions by the Council of Economic Advisers and the Congressional Budget Office is reflected in similar estimates of the aggre- gate impact of the stimulus package. The Congressional Budget Office (2010) estimates that, in the first quarter of 2010, the stimulus package raised the level of GDP by between 1.7 percent and 4.2 percent and raised the level of employment by 1.2 million to 2.8 million. The CEA (2010) recently estimated that, by the second 7 This controversy was highlighted by a Wall Street Journal article describing a poll of professional forecasters. When asked about the net effect of the stimulus on economic growth and employment, 38 of the forecasters answered that ARRA had a positive effect, while six answered that the stimulus package had a negative effect (Izzo, 2010).